A new algorithm may improve image stitching, resulting in better panoramas
The Society for Industrial and Applied Mathematics published a paper written by a couple of math professors who propose a new mathematical models for image stitching, which should improve the image blending when creating panoramas.
Panoramic photography is popular, the science behind it is hard
Panoramic images have become popular as there are many consumer cameras and even mobile phones offering such built-in functionality. Still, this is a relatively new area with a lot of room for improvement left.
A panoramic image is made from a number of at least two smaller images, assembled together, operation called image stitching. The stitching happens in two stages: first the images are analyzed for some common points in the overlapping area and this is the image alignment area. The second stage, after the images are aligned, is the image blending, when the components are combined seamlessly. If each of the initial images was taken in different conditions, for example if the camera was put in an Auto mode, they can vary largely in exposure, light and colors, so a transition area may be seen in the resulting panorama.
Math professors research and improve the science behind assembling panoramas
A couple of math professors, Wei Wang at Tongji University in Shanghai, China and Michael Ng is at Hong Kong Baptist University authored a paper called ‘A Variational Approach for Image Stitching I’ which was published by The Society for Industrial and Applied Mathematics (SIAM). They address the image blending part of the stitching, proposing a new approach.
The two professors propose a variational approach to the stitching process containing an energy functional to determine both a weighting mask function and a stitched image together. They also present and study a novel color correction procedure to deal with color inconsistencies of different images in the stitching process.
Better panoramas and further: making use of degraded images, extending the use in the medical and video fields
The researchers are optimistic because their model may handle images which are degraded or have localized missing parts or occlusions. Providing some other clear images which are similar in content with the target panorama, it should be possible to stitch together the damaged images using the proposed mathematical model.
As any university research, this is only a step which can be improved by further research. There is a possibility that by extending this variational approach, it would be able to handle three-dimensional image stitching which is used in medical imaging applications and also stitching video in computer applications.